One method of reproducing an omnidirectional image involves mapping the image to a 3D model for perspective projection thereby to generate and display a display image in the visual field range in a viewer's visual line direction. The omnidirectional image is obtained by having 360-degree surrounding images in the horizontal direction and 180-degree surrounding images in the vertical direction mapped onto a 3D model.
This method uses only part of the omnidirectional image in generating the display image. For example, in the case where the viewing angle in the horizontal direction is 90 degrees, only a quarter of the entire omnidirectional image is used in the horizontal direction to generate the display image. This applies to the vertical direction in similar way. That region of the omnidirectional image which is used to generate the display image is thus small in proportion with respect to the whole omnidirectional image.
However, because a viewer's visual line direction is designated by the viewer at the time of reproduction, it is difficult for a recording apparatus to record only that portion of the entire omnidirectional image which will be used to generate the display image. For this reason, it has been desired that the amount of omnidirectional image data to be transmitted from the recording apparatus to a reproduction apparatus be reduced by having the entire omnidirectional image recorded on the recording apparatus and by allowing the reproduction apparatus to request the recording apparatus to send only the image region to be used to generate the display image.
Therefore, it has been proposed that the omnidirectional image be divided into multiple regions each of which is turned into an encoded stream so that only the encoded stream of the desired divided region is transmitted from the recording apparatus to the reproduction apparatus (e.g., see PTL 1).
It has also been proposed that an entire omnidirectional image at low resolution be encoded separately from divided and partially overlapping regions of an omnidirectional image at high resolution, and that an encoded stream of the entire low-resolution omnidirectional image and an encoded stream of only that divided region of the high-resolution omnidirectional image which corresponds to the viewer's visual line direction be transmitted from the recording apparatus to the reproduction apparatus (e.g., see PTL 2).
In the preceding case, of the divided regions of the high-resolution omnidirectional image, only the one used to generate the display image needs to be transmitted. This translates into a reduced amount of transmitted data compared with the case where all divided regions of the high-resolution omnidirectional image are transmitted. It is also possible to generate the display image using the low-resolution omnidirectional image in the case where a divided region other than the divided region of the high-resolution omnidirectional image corresponding to the transmitted encoded stream is used to generate the display image or where the viewer's visual line direction is abruptly changed. Further, in a case where the omnidirectional image is divided into numerous regions or in a case where regions spanning the edges of the omnidirectional image are set to be part of the divided regions, it is possible to increase the proportion, with respect to the entire omnidirectional image, of the high-resolution omnidirectional image used to generate the display image.